Group Presented Two Papers at IEEE ICASSP 2026
Our group presented two research papers at the IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) 2026, held in Barcelona, Spain. These works cover advancements in X-ray computed laminography and mm-wave radar signal processing.
1. LamiGauss: Sparse-View X-Ray Laminography Reconstruction
LamiGauss addresses the critical challenge of reconstructing high-quality 3D volumes from sparse-view laminographic projections. This is essential for non-destructive inspection of plate-like structures, such as microchips and composite battery materials, where traditional CT is geometrically constrained.
Our method combines Gaussian Splatting radiative rasterization with a dedicated detector-to-world transformation model. By leveraging a novel initialization strategy that filters out common laminographic artifacts, LamiGauss concentrates model capacity on genuine structures. It achieves superior reconstruction quality using only 3% of full views compared to traditional iterative methods.
Reference: C. Chen et al., “LamiGauss: Pitching Radiative Gaussian for Sparse-View X-Ray Laminography Reconstruction,” in IEEE ICASSP 2026. DOI: 10.1109/ICASSP54000.2026.11460905
2. DOA Estimation for Mm-Wave FMCW Radar
This research introduces an accurate and efficient Direction-of-Arrival (DOA) estimator for mm-wave FMCW radar that requires only two receiving antennas. Conventional methods often require larger antenna arrays to achieve high accuracy.
Our approach applies an exact formula to super-resolve the beat frequency and phase of radar signals. We validated the estimator through extensive numerical simulations and real-world radar measurements, achieving a maximum DOA error of no more than 0.6°. This work significantly reduces hardware complexity while maintaining high-precision sensing capabilities.
Reference: Q. Dai et al., “DOA Estimation for Mm-Wave FMCW Radar with Two Receiving Antennas,” in IEEE ICASSP 2026. DOI: 10.1109/ICASSP54000.2026.11460696